When researchers want to understand how a drug changes the body, they look beyond the scale. Total weight is a blunt tool. What matters more is the mix inside that number: how much is fat, how much is lean tissue like muscle, and where the fat is actually sitting. A large 2026 systematic review published in Diabetes, Obesity and Metabolism set out to answer exactly that question for a class of peptides called glucagon-like peptide-1 receptor agonists, or GLP-1 receptor agonists.
The research team pulled data from 43 randomised controlled trials covering 3,379 participants. Those participants were adults living with overweight or obesity, some with type 2 diabetes and some without. The analysis compared 17 different GLP-1 based interventions or dosages against a control group, making it one of the most comprehensive head-to-head comparisons of these peptides on body composition to date.
The headline finding cuts two ways. On the encouraging side, every subcutaneous GLP-1 receptor agonist studied outperformed the control group at reducing total body fat, fat mass, visceral fat, subcutaneous fat, and liver fat. On the concerning side, several specific doses, particularly at the higher end, were linked to measurable reductions in lean mass. That tradeoff is now drawing serious attention in the research community.
What the researchers measured
Body composition is not a single number. The review tracked seven distinct outcomes: total body fat expressed as a percentage, fat mass in kilograms, visceral adipose tissue area, subcutaneous adipose tissue area, liver fat content as a percentage, total lean tissue as a percentage, and lean mass in kilograms. Measuring this many variables at once gives a much richer picture than weight alone.
Visceral adipose tissue, often called VAT, is the fat that wraps around the organs inside the abdomen. Research has long linked high VAT to elevated cardiovascular and metabolic risk, so it receives particular attention in obesity studies. Subcutaneous adipose tissue, or SAT, is the fat sitting just under the skin. Liver fat, sometimes called hepatic steatosis when it accumulates in excess, is a separate concern tied to metabolic liver disease.
To combine results from trials that used different measurement tools and different scales, the researchers expressed all findings as standardised mean differences, a statistical method that puts every result onto a common ruler. They also rated how confident they were in each finding using an established framework called Confidence in Network Meta-Analysis.
Fat reduction across the board
On the fat side of the ledger, the results were consistent. Every subcutaneously administered GLP-1 receptor agonist in the dataset produced statistically significant reductions in total body fat, fat mass, visceral adipose tissue, subcutaneous adipose tissue, and liver fat compared with the control group. That consistency across 43 trials and multiple peptide variants strengthens confidence in the signal.
Visceral fat reductions are particularly notable from a research standpoint. Because VAT is metabolically active tissue that secretes inflammatory signals, the literature treats its reduction as a meaningful physiological change, not just a cosmetic one. The review documented significant VAT area reductions across the GLP-1 class as a whole, though the magnitude varied by peptide and dose.
Liver fat reductions were also observed across the class. A recent abstract from this same dataset reported significant decreases in liver fat content from baseline, which aligns with earlier smaller trials that had examined individual agents in isolation. The network meta-analysis design allowed the researchers to compare those earlier findings in a more structured way.
The lean mass concern
The finding that attracted the most discussion is the lean mass data. When the researchers looked at total lean tissue as a percentage of body mass, no statistically significant differences appeared between the GLP-1 groups and the control group. That sounds reassuring at first glance.
However, when lean mass was measured in absolute kilograms rather than as a proportion, a different picture emerged at certain doses. Three specific regimens produced statistically significant lean mass reductions from baseline: one peptide at 1.8 milligrams per day delivered subcutaneously daily, a second at 1.0 milligram weekly, and a third at 15 milligrams weekly. The standardised mean differences for those three ranged from negative 0.50 to negative 1.09, which in statistical terms represents a moderate to large effect size.
Why the discrepancy between lean tissue percentage and lean mass kilograms? When overall body mass drops sharply, lean tissue can appear stable or even increased as a proportion of total mass, even if the absolute amount of lean tissue is shrinking. Researchers flag this as a measurement artifact that can obscure real muscle loss if only percentage figures are reported. The review's use of both metrics is one of its methodological strengths.
Dose and delivery route
All the results described above apply specifically to subcutaneous delivery, meaning injections under the skin. The review focused on this route because the trials in the dataset predominantly used it. Oral formulations were addressed separately and did not show the same pattern across all outcomes, though the number of oral-route trials was smaller.
Dose appeared to matter for lean mass outcomes in particular. The three regimens associated with significant lean mass reduction were all at the higher end of the dosing ranges studied. Lower-dose comparators in the same peptide families did not always reach statistical significance for lean mass loss, suggesting a dose-response relationship worth tracking in future research.
The literature suggests this may partly reflect the magnitude of overall calorie deficit these peptides produce. A larger energy deficit, regardless of its cause, is generally associated with some degree of lean tissue loss alongside fat loss. Whether the lean mass reduction seen here is driven by calorie restriction, a direct peptide effect, or both remains an open question the review could not fully resolve from existing trial data.
Network meta-analysis as a method
A traditional meta-analysis pools results from trials that share the same direct comparison. A network meta-analysis goes further by combining both direct and indirect comparisons, allowing researchers to rank interventions even when no single trial tested them head to head. This review used a frequentist network approach with a random-effects model, which is a standard and conservative choice that accounts for genuine variation between trials.
The 43 trials covered 17 different interventions or dosages, and the search ran across six databases from inception through November 2025 with no language restrictions. That breadth reduces the risk of cherry-picking only English-language or positive-result studies. The team also searched grey literature, meaning unpublished or non-journal sources, to limit publication bias.
The certainty of evidence was rated using the Confidence in Network Meta-Analysis framework, often abbreviated CINeMA. This framework grades how much trust readers should place in each finding, accounting for factors like trial risk of bias, inconsistency between trials, and how precise the estimates are. The review reported these ratings alongside its main results, which is important context for interpreting the findings.
What the research leaves open
A few important questions sit beyond what this review could answer. First, the trials varied in length, and most were not designed to track long-term lean mass trajectories. Whether the lean mass reductions observed are permanent or partly reversible with time or resistance exercise is not clear from this dataset.
Second, the clinical significance of the lean mass changes measured here is still debated. A statistically significant reduction in absolute lean mass kilograms does not automatically translate to a clinically meaningful loss of muscle function. Trials that also measured physical performance, grip strength, or functional assessments alongside body composition would give a clearer picture, and most of the trials in this dataset did not.
Third, the review grouped participants with and without type 2 diabetes together in its main analyses. Whether the body composition response differs meaningfully between those subgroups is a relevant question, particularly given that type 2 diabetes itself alters muscle metabolism. The authors acknowledged this as a limitation. Early data points toward subgroup analyses being a productive area for future randomised trials in this space.
